xb95 filled in on a few slides, on the stuff he'd worked on.Less questions than Seattle, but I think the slides/info were more complete/dense this time around, too. It was nice to have a lot of the stuff working that was kinda hand-wavvy when I was in Seattle.
In particular, persistent connections and connection verification.
When I was in Seattle somebody asked, "But, with all this work you're doing to squeeze every drop of performance, why's the site still slow?" I just kinda cried inside, knowing it'd get better.
Once we got persistent connections going and could do perfect load balancing, the webs were never idle when the site was slow and jobs just flowed perfectly (as fast as possible) without TCP fighting and doing timeouts.
It was nice to be able to ssh into livejournal's servers and show them the state of all 6 perlbal processes, all totally happy and unloaded, along with unloaded web nodes.
We've come far. Can't wait to "solve" the DB problem as well. That's making progress, almost as fast as I'd like.
I think it's time to start documenting Perlbal more and giving it a website/mailing-list/etc.
After the talk was beer with a lot of Perl geeks. Good conversations going on. Nice to be around a lot of smart people, and net.famous people too. It doesn't shock me as much as it used to, but it's still nice to meet new people. Whitaker's amused whenever net.famous people are around. He was like, "Tell me who else was there! Anybody I'd know?" I replied, "Well, that Randal guy with the maroon shirt talking about the party.... he co-wrote your Learning Perl book."
Dina's out watching Napolean Dynomite now. I saw it in California the other week.
What to do, what to do...
Think I broke my NES emulator on the xbox. It hangs loading it now. Something corrupt? Should recopy from my brother's.
The end.